_____________________________
Urban Rapid Rail Transit and
Gentrification in Canadian Urban
Centres – A Survival Analysis
Approach
Annelise Grube-Cavers
Zachary Patterson
January 2013
CIRRELT-2013-06
Bureaux de Montréal :
Bureaux de Québec :
Université de Montréal
C.P. 6128, succ. Centre-ville
Montréal (Québec)
Canada H3C 3J7
Téléphone : 514 343-7575
Télécopie : 514 343-7121
Université Laval
2325, de la Terrasse, bureau 2642
Québec (Québec)
Canada G1V
G1V0A6
0A6
Téléphone : 418 656-2073
Télécopie : 418 656-2624
www.cirrelt.ca
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres A Survival Analysis Approach
Annelise Grube-Cavers, Zachary Patterson*
Interuniversity Research Centre on Enterprise Networks, Logistics and Transportation (CIRRELT)
and Department of Geography, Planning and Environment, Concordia University, 1455 de
Maisonneuve W., H 1255-26 (Hall Building), Montreal, Canada H3G 1M8
Abstract. Despite the existing knowledge that urban rapid rail transit has a multitude of
effects on surrounding areas, and despite some attempts to understand the links between
transit and gentrification, there has yet to be a systematic study of if, and how, proximity to
transit plays a role in transforming neighborhoods. This research addresses the topic by
assessing the relationship between the implementation of urban rapid rail transit and
gentrification using an application of the statistical technique survival analysis. It
specifically tests whether proximity to rail transit is related to the phenomenon of
gentrification in Census Tracts (CTs) in Canada’s three largest cities. It is found that
proximity to rail transit, as well as proximity to other gentrifying census tracts, have a
statistically significant effect on gentrification in two of the three cities analyzed. By
providing a methodological framework for the empirical analysis of the impact of urban rail
transit on gentrification, this paper is a reference for both researchers and transportation
planners.
Keywords. Gentrification, urban transit, rapid rail transit, survival analysis, urban
development, neighborhood change.
Acknowledgements. This research has been funded through the Social Sciences and
Humanities Research Council of Canada (SSHRC) Canada Research Chair’s program as
well as the Joseph-Armand Bombardier Canada Graduate Scholarships Program. In
addition, the authors would like to extend a special thanks to: Craig Townsend, Alex
Guindon and Robert Tomaszewski at Concordia University, Nathalie Vachon and Dany
Fougères at the Institut National de la Recherche Scientifique (INRS), and Marcel Fortin at
the University of Toronto’s Computing in the Humanities and Social Sciences (CHASS) as
well as Richard Maaranen and the Centre for Urban and Community Studies for providing
access to digitized census data and digital boundary files. Their help was essential to the
progress of this research. Special thanks to Chris Harding and Arash Mostofi for their
support.
Results and views expressed in this publication are the sole responsibility of the authors and do not
necessarily reflect those of CIRRELT.
Les résultats et opinions contenus dans cette publication ne reflètent pas nécessairement la position du
CIRRELT et n'engagent pas sa responsabilité.
_____________________________
* Corresponding author: [email protected]
Dépôt légal – Bibliothèque et Archives nationales du Québec
Bibliothèque et Archives Canada, 2013
© Copyright Grube-Cavers, Patterson and CIRRELT, 2013
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
INTRODUCTION
Transit is widely recognized to have effects beyond increasing accessibility. These effects,
however, are not fully understood. The relationship between good accessibility to transit and the
gentrification of surrounding neighborhoods, for instance, has recently been brought up in planning
literature, but surprisingly little research examines this relationship explicitly. Attempts to
investigate the link have suffered from two principal weaknesses. First, previous research has used
definitions of gentrification that are not consistent with the extensive literature on the subject.
Second, previous literature has not adequately accounted for the fact that gentrification is not an
instantaneous phenomenon, but one that takes place over long periods of time.
The present research addresses the question, how does the implementation of transit
networks, and more specifically urban rapid rail transit, affect the onset of gentrification in
Canadian metropolitan areas over time. This study differs from publications in this burgeoning field
by: adopting a definition of gentrification that is consistent with the majority of the literature; and
explicitly incorporating time (until gentrification, as well as time exposed to transit) through the use
of the statistical technique, Survival Analysis. This approach is used to test the relationship between
proximity to urban rapid rail transit and gentrification in the three largest Canadian cities.
The metropolitan areas of Toronto, Montreal, and Vancouver have evolved in a variety of
cultural, economic and geographical contexts resulting in markedly different patterns of
development. Nevertheless, they are evaluated in this study using a consistent method of analysis in
order to provide comparable results and gain insight into potential urban rail transit and
gentrification linkages.
The questions addressed, and the results found, in this paper are important in providing
guidance to researchers and planners as they seek to provide equitable and accessible transit. The
methodology and outcomes of this research will function as tools to evaluate the potential for
negative effects that can occur as a result of infrastructure investment.
CIRRELT-2013-06
1
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
The paper starts with a review of the existing literature, followed by a description of the
cities analyzed, and the data and methodology used to establish which census tracts (CTs) are
susceptible to gentrification and which have gentrified. Survival Analysis, is then introduced and
described and finally, results from model estimation and a discussion of those results are presented.
LITERATURE REVIEW
Three aspects of the gentrification literature are important to understand the context of the
research presented in this paper. First, a basic overview of the general themes of gentrification is
needed to provide context. Second, a description of how gentrification is typically conceived of and
operationalized is given. Third, more detail on the transit-gentrification literature is provided in
order to identify existing weaknesses in this field of research.
Gentrification: A Large Literature
There have been hundreds of papers written on the topic of gentrification since the term
was first defined by British Sociologist, Ruth Glass in the early 1960s (Glass, 1964). While the term
has been applied to many different contexts it has been associated with central city, urban
neighbourhoods and typically involves two core elements. The first element is the notion that a
gentrifying neighbourhood is one that was previously poor, or ‘working class.’ The second is that
the social status of the neighbourhood, (i.e., income, education and percentage of residents in
professional occupations,) is increasing faster within the neighbourhood than the city as a whole,
this is associated with changes in the housing market (i.e., rising rents and house values).
The literature seeking to describe the phenomenon of gentrification can be characterized by
its theoretical approach and the particular analytical methods used. While various theoretical
classifications could be used, one common distinction is that made between ‘supply-side’ and
‘demand-side’ perspectives. Supply-side proponents believe that gentrification is driven by policies
intended to revitalize ultimately drawing ‘gentrifiers’ into neighborhoods (Slater, 2006). Demand-
2
CIRRELT-2013-06
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
side proponents, on the other hand, believe that gentrification is caused by changes in societal
preferences – whereas residents previously sought to flee the city for the suburbs, gentrification
results from the increased desirability of central-city neighbourhoods (Ley, 1986).
With respect to the analytical methods adopted the literature can also be categorized as
either qualitative or quantitative. Quantitative or empirical analyses of the process of gentrification
have tended to use census statistics to define and explain gentrification (Pollack, Bluestone and
Billingham, 2010), whereas qualitative approaches focus primarily on interviews and observations
(Rose, 2004; Newman and Wyly, 2005).
Throughout the literature, from both the supply- and demand-side perspectives, the link
between accessibility to transit and gentrification has been mentioned (Filion, 1991; Atkinson and
Bridge, 2005; Skaburskis and Mok, 2006; Walks and Maaranen, 2008). Despite this relatively
common statement there has been surprisingly little research to have looked at this question
specifically. The research that does broach this topic, further elaborated upon below, seems to have
overwhelmingly taken a quantitative approach. Before describing this literature in more detail, it is
first necessary to look at how gentrification has been operationalized in the quantitative literature on
the topic.
Identifying Gentrification
In reviewing the literature it is clear that a concensus has developed amongst the
quantitative studies about how to identify gentrification. First, quantitative analyses of gentrification
use neighborhoods as the unit of analysis, and census tracts are generally used as a proxy for
neighborhoods (Kahn, 2007; Freeman, 2005; Walks and Maaranen, 2008; Pollack, Bluestone and
Billingham, 2011). Second, as suggested by the description of gentrification outlined in the previous
section, there are two important criteria of gentrification.
The first criterion for gentrification is that in order for a neighborhood to undergo
gentrification, it must be considered a ‘working class’ neigborhood at the beginning of the period of
CIRRELT-2013-06
3
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
analysis - in other words, to be included in the analysis of gentrifcation, a neighborhood needs to
first be ‘gentrifiable’ (Freeman, 2005; Walks and Maaranen, 2008). While different authors have
operationalized this in different ways, typically, neighborhoods are considered to be gentrifiable if
the average, or median, income of the census tract is below the average of the Census Metropolitan
Area (CMA) within which it is found (Freeman, 2005). Other indicators of social status are also
sometimes used – for example Walks and Maaranen (2008) employ not only income in their
classification of ‘working class’ neighbourhoods, but also other variables, including levels of
educational attainment. The second part of recognising gentrification entails observing changes in
‘status’ of the neighbourhood in question. Gentrification is defined as a process whereby the middle
and upper classes move into ‘gentrifiable’ neighbourhoods.
While the identification of census tracts considered to be gentrifiable is relatively
straightforward, empirically distinguishing gentrification in an area is more complex. In addition to
characteristics of the residents themselves - including higher levels of educational attainment,
higher incomes, and an increasing number of professionals - , some characteristics of the housing
stock, such as housing values or rents are also considered, since these are important indicators of the
gentrification of neighborhoods. Analysis of gentrification is typically carried out by measuring
whether all of the relevant indicators (see section on Data for more details) improve at a rate faster
than for the CMA as a whole (Freeman, 2005). Critical to this then is the notion that gentrification
is a relative process, gauged against changes in the CMA being analysed, and is identified by the
use of several indicators jointly.
Transit and Gentrification
With this background established, it is now possible to discuss the literature that explicitly
tests the link between transit and gentrification. It is worth noting that a significant number of
studies have focused on the relationship between transit and housing values, a common indicator
used in studies of gentrification (Hess and Almeida, 2007; Atkinson-Palombo, 2010; Cervero,
4
CIRRELT-2013-06
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
Bernick and Gilbert, 1994). Despite these studies and observations in the gentrification literature,
the relationship of transit and gentrification has remained largely overlooked. The three studies that
come the closest to explaining this link are described below.
In a study of Northeastern Chicago, Lin equated higher housing values near transit stations
to the presence of gentrification (Lin, 2002). Although, as mentioned above, increased housing
values is an important indicator of gentrification, housing costs alone are generally considered in
conjunction with other indicators to identify gentrification.
A second study to examine the relationship of transit and gentrification found that
neighborhoods in 12 urban centres that had transit implemented between the years 1990 and 2000
also showed gentrification (Pollack, Bluestone and Billingham, 2010). Indicators typically
associated with gentrification are used in their analysis, however these indicators are not considered
jointly and no attempt is made to identify neighborhoods that would be considered gentrifiable. The
study period for this paper was only ten years, and did not take into account how long transit had
been present for (Pollack, Bluestone and Billingham, 2010).
The final of these three papers examining the transit-gentrification linkage, is by Matthew
Kahn (2007). The methodology and results of this paper leave many unanswered questions. The
author, using regression analysis, estimates the impact of rail transit on income and the proportion
of college graduates living in an area (Kahn, 2007). As with Lin’s (2002) paper, Kahn (2007)
omitted a number of widely cited indicators of gentrification including professional occupations
(Filion, 1991; Walks and Maaranen, 2008). Moreover, the indicators included in the study (income,
proportion of college graduates), are considered independently of one another. Unlike the two
papers described earlier, Kahn (2007) assesses his study areas over a longer period of time, and also
accounts for the length of time that census tracts are exposed to transit. Ultimately the author
concluded that while some walk-and-ride stations did seem to bring about gentrification, park-andride stations did not (Kahn, 2007).
CIRRELT-2013-06
5
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
By not acknowledging the dynamic nature of gentrification, current literature is inconsistent
in both descriptions and definitions. For a study’s methodology to be consistent with the existing
literature, the gentrifiability of an area must first be assessed. There must also be adequate
justification given for the inclusion of indicators, and they must be assessed jointly in a given area
to establish the presence of gentrification.
There are notable methodological omissions that appear in the key references on transit and
gentrification. The first is the lack of analysis to establish whether or not the neighborhoods
included are actually susceptible to gentrification, or ‘gentrifiable.’ Second, none of the transitgentrification studies identify gentrification using multiple variables jointly. Two of the three
reference papers fail to adequately take into account the temporal nature of the process of
gentrification, an aspect that should be central to the methodologies used. Additionally, evaluating
the effect of transit presence as instantaneous limits the scope of the analysis, such that long-term
effects of transit infrastructure may not be captured.
Building upon the work described above, the research presented here uses a systematic
approach to identify gentrifiability and gentrification, while at the same time explicitly examining
the impact on gentrification of the proximity to transit stations as networks are implemented. This
research explicitly accounts for the dynamic nature of the phenomenon by employing an innovative
application of survival analysis to test the significance of proximity to rail transit, which may
change over time as networks are expanded, to the onset of gentrification.
STUDY AREAS – MONTREAL, TORONTO AND VANCOUVER
The study areas included in this research are the CMAs of Montreal, Toronto and
Vancouver; the three largest urban centers in Canada (see Figure 1 below). Toronto is the largest
Canadian city with a total CMA population of 5.1 million residents in 2006 (Statistics Canada,
2006). Its rapid rail transit system, the Toronto Subway, opened its first stations in 1954 and in 2006
6
CIRRELT-2013-06
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
had a total of 69 stations extending 70 kilometers through the City of Toronto (Toronto Transit
Commision, 2012).
Montreal, located in the province of Quebec, is the next largest Canadian city with a CMA
population of 3.5 million (Statistics Canada, 2006). The Montreal Metro inaugurated its first
stations in 1966 (Clairoux, 2001). As of 2006 it consisted of 68 stations on the island of Montreal
which were built and opened in 11 increments between 1966 and 1988. Only on-island CTs were
analyzed for Montreal as the water barrier of the St Lawrence River was expected to eliminate any
relationship between the onset of gentrification and the presence of transit. Additionally, the offisland metro stations were excluded because of the dates of inauguration (the three newest stations
to the north opened in 2007) and drastically altered census tract boundaries, surrounding the stations
to the south of Montreal, in Longeuil.
The third (and third largest) city is Metro Vancouver and, as of 2006, it had a significantly
smaller CMA population of 2 million (Statistics Canada, 2006). Vancouver’s Skytrain opened its
first stations in 1986 for the Vancouver World Exposition. The system has since expanded, and in
2006 had 32 operational stations along 68.7 kilometers of track. It provides service to the city of
Vancouver as well as four adjacent municipalities which are all a part of Metro Vancouver (Metro
Vancouver, n.d.). An additional line has since opened, but the stations came into operation after the
end of the study period.
Defining the boundaries of the study for Vancouver was more challenging than for the other
cities since there was no obvious delineation that encompassed the whole Skytrain system. Walks
and Maaranen analyzed only the City of Vancouver proper, but that geography excludes the
majority of transit stations and was therefore deemed inappropriate for the scope of this study.
Three conditions were used to establish the study area boundaries: first, any CTs separated
from the Skytrain by water were eliminated. Second, population density had to be in the 90th
percentile to be considered a part of the ‘urban’ area, as gentrification has previously been defined
CIRRELT-2013-06
7
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
as a phenomenon observable in urban neighborhoods (Ley, 1986; Freeman, 2005). The census tracts
were also excluded if they were not a part of the contiguous area of ‘urban’ CTs, that is, if they
were separated from the Skytrain by areas with low population density.
FIGURE 1: Study areas
DATA USED
To establish whether the process of gentrification is occurring, empirical studies use census
data to distinguish changes in the makeup of neighborhoods (Filion, 1991; Freeman, 2005; Walks
and Maaranen 2008). Specific indicators of gentrification used in the past to measure the process
include demographic statistics such as: population; household, family and individual income;
college or university education levels; persons employed in professional occupations; household
structure (referring to the number of children in a household); and racial and ethnic composition
(particularly in the USA). Indicators related to housing and location are also taken into account,
typically as statistics on the number of housing units, housing tenure and age of the housing stock;
housing costs, both the value of homes and costs of rent; distance from CBD; distance from other
8
CIRRELT-2013-06
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
urban amenities, including parks as well as commercial districts (List compiled from; Filion, 1991;
Freeman, 2005; Walks and Maaranen, 2008; Pollack, Buestone and Billingham, 2011).
Data used in the study originated from Statistics Canada and was aggregated to the CT
level. Census tracts are small geographical areas delineated by Statistics Canada that have an
approximate population of 2,500 to 8,000 residents and that are supposed to remain relatively stable
over time (Statistics Canada, 2007). Despite this supposed stability, one challenge was rendering the
census tract boundaries comparable across census years. The most common change in census tracts
over time is for them to split into smaller census. In this study the authors decided to aggregate
adjacent CTs that had undergone changes in order to make the 2006 census tract boundary file
comparable to that of the first year in each of the study periods. Although the results of the analysis
are less fine-grained, they are more accurate than if we were to normalize the boundaries to later
years.
Study periods for the cities vary depending on when their respective transit systems first
came into operation. For Montreal the first year of census data used is 1961, five years before the
first stations opened. In Vancouver the study period only begins in 1981 as the Skytrain was first
inaugurated in 1986. The case of Toronto is more complicated as comparable data at the CT level
was not available for 1951, the desired year since the subway opened in 1954. The literature on
gentrification in Toronto identifies it as a process that was not really observed until the 1970s and
1980s (Walks and Maaranen, 2008), as such the study period for Toronto is the same as Montreal:
1961-2006. Missing data in 1966 and 1976 led them to be excluded from the analysis for both
Montreal and Toronto so the full data set includes census years 1961, 1971, 1981, 1986, 1991,
1996, 2001 and 2006.
In addition to census statistics, other data was collected in an effort to test, statistically, the
relationship between aspects of the urban environment and gentrification. This data included the
CIRRELT-2013-06
9
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
distances from the centroids of each census tract to the nearest transit station for every year from the
time that the first stations were opened. Past studies have explored the idea that gentrification is
related to the distance from the Central Business District (CBD) (Lin, 2002; Kahn, 2007; Filion,
1991). To incorporate this idea, the distance was also measured from the centroids of census tracts
to the centroid of the CBD, as defined by the delineations offered on each of the CMA government
websites. Another condition mentioned in the literature is the presence of older housing stock with
the architectural character desired by ‘gentrifiers’ (Filion, 1991; Walks and Maaranen, 2008; Ley
1986). Age of housing stock was defined by the proportion of housing that was constructed before
1946; this was drawn from past emphasis put on pre-WWII inner-cities (Walks and Maaranen,
2008).
The distance was also measured from the centroid of each census tract to the nearest large
park (defined as any park equal to or exceeding 50 000 square metres) or major body of water (lake,
river or ocean). Many authors refer to the urban amenities that draw gentrifiers to the central city
neighbourhoods. Often urban amenities refer to commercial districts, or areas of consumption that
cater to middle and upper income residents (Smith and Williams, 1986). Since these urban attributes
are impossible to accurately track over time they were excluded from this analysis, but both parks
and proximity to water, viewed as urban amenities (Helms, 2003) which enhance the value of a
location were included to test their statistical relevance to gentrification. The final variable that was
added was the distance from the centroid of each census tract to the centroid of the nearest CT that
had experienced gentrification. This has not previously been considered in the literature, but was
suggested as a potentially important variable when preliminary results of this research were
discussed.
10
CIRRELT-2013-06
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
METHODOLOGY
The following two subsections explain the methodology: the identification of gentrification and the
survival analysis.
Identifying Gentrification
In order to conduct a statistical analysis of the effect of urban rail transit on gentrification, it is
necessary to identify the CTs that could be considered gentrifiable and those that have actually
experienced gentrification over the study period.
To establish whether a CT was gentrifiable the average family income of a census tract, and the
number of degrees per capita were assessed, both of which needed to be lower than the CMA
average for a given census year. If this was the case for a CT, then the CT in question was included
in the sample for the statistical analysis. The full list of indicators used in the identification of the
onset of gentrification is below;

average monthly rent,

proportion of people in professional occupations living in a census tract,

percentage of owner-occupied dwellings,

average family income

and number of degrees per capita.
For a CT to be considered to be gentrifying, all of the aforementioned indicators had to have
experienced a change greater than the change experienced in those indicators for the CMA, as well
as being considered gentrifiable in the first census year of the census period (i.e. to have gentrified
in 1971, it must be considered gentrifiable in 1961, the preceding census year in the dataset).
Indicators used to establish gentrification were limited to those that were available for census years
throughout the study period. Table 1 below summarizes how many CTs were included for each
CIRRELT-2013-06
11
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
CMA, how many of those were considered gentrifiable according to our criteria, and how many
ultimately gentrified.
TABLE 1: Gentrifiable and Gentrifying Tracts for All Cities
City
CTs Included in Study
Gentrifiable CTs
Gentrifying CTs
Montreal
291
220
86
Toronto
248
165
71
Vancouver
143
101
19
FIGURE 2: Gentrifiable and gentrifying tracts in Montreal (1961-2006)
Figure 2, above, shows the results of the gentrification analysis for Montreal. It is worth
noting that the southernmost metro-line running North-West to South-East has very few gentrifiable
CTs around it. Overall there appears to be an association between the onset of gentrification and the
12
CIRRELT-2013-06
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
presence of metro stations. Although there are some outlying tracts that appear to have experienced
the demographic changes defining gentrification, there is a noticeable pattern of the effect in areas
accessible to the metro.
FIGURE 3: Gentrifiable and gentrifying tracts in Toronto (1961-2006)
The gentrification in Toronto shadows the Toronto Subway system in a similar, if not more
striking, way than the process visible in Montreal. As seen in Figure 3 above, the gentrification
occurring in Toronto is focused along two axes, one running North South along the westernmost
branch of the Yonge-University-Spadina line, and one running East-West along the Bloor-Danforth
line. Many of the most commonly cited gentrifying areas of Toronto are captured in this analysis
(Slater, 2004; Hackworth and Rekers, 2005; Murdie and Texeira, 2011).
As the smallest area included for analysis, Vancouver had only 143 CTs included in the
analysis (see Table 1 for details). Figure 4, below, depicts the gentrifiable and gentrifying tracts in
Vancouver. The CTs that had undergone some form of gentrification in Vancouver are significantly
CIRRELT-2013-06
13
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
sparser than in the other two CMAs and shows notably less of a pattern in terms of their overall
dispersal through the CMA.
FIGURE 4: Gentrifiable and gentrifying tracts in Vancouver (1981-2006)
Survival Analysis
This section describes the statistical methods and results used in this study. The approach outlined
analyzed the relationship between the presence of rapid rail transit and gentrification in Canadian
urban centers. The statistical technique is one that was used in an earlier paper by the authors,
presented at the 91st annual meeting of the TRB (Author B and Author A, 2011). As in the previous
paper, the dependent variable in this analysis is binary since CTs will either have experienced
gentrification or not, as such a limited dependent variable statistical method needed to be used
(Kennedy, 2003; Maddala, 1983). Since this research project integrates multiple different urban
centers the analysis was complex and needed to be adapted to be suitable for all cases included.
14
CIRRELT-2013-06
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
Primarily used in the field of bio-statistics, Survival Analysis has had limited use in transportation
research, but it is still particularly relevant to the field (Washington, Karlaftis and Mannering,
2003). Survival Analysis is defined as “a collection of statistical procedures for data analysis for
which the outcome variable of interest is time until an event occurs” known as the survival time
(Kleinbaum and Klein, 2005). This type of analysis is particularly useful when working with
variables whose effects vary over time, called time-dependent variables. For various applications of
survival analyses see the text written by Kleinbaum and Klein (Kleinbaum and Klein, 2005).
In the model presented in this paper the ‘event’ in question is gentrification and the ‘treatment’
variable would be considered the presence of rapid rail transit close to the census tract. Thus, the
results of the model give the survival time of census tracts until they gentrify based on the presence
of transit over time as well as a few other variables defined below.
As was discovered in a preceding paper (Author B and Author A, 2011), the presence of transit was
more difficult to integrate than predicted. Preliminary analyses showed that there does not appear to
be a linear relationship between gentrification and the independent variable of proximity to transit.
As a result, a gravity function was used to capture the effect of distance from transit to an individual
CT. The gravity measure is calculated for each census tract for each year of the study as a function
of ‘cdist’ as shown in Equation 1 below (for more details on the calculation of gravity measures see
de Dios Ortúzar and Willumsen, 2001). Different betas were used for each city after testing to see
which had the most statistically significant influence on the model.
Eqn. (1)
The variable of proximity to transit, or ‘exposure’ as the gravity measure is called is a time
dependent variable meaning that the variable changes over time. The distance to transit stations,
‘cdist’, may change every time new transit stations are added to the transit network. Before being
CIRRELT-2013-06
15
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
included in the statistical analysis, the variable is normalized to one to facilitate comparability of
coefficients across cities.
THE SURVIVAL MODEL RESULTS
Introduction to Estimating Survival Models
Following the example of the preceding paper which included only preliminary analysis for
Montreal, the Extended-Cox (EC) model was selected for this statistical analysis (Author B and
Author A, 2011). The EC model is a semi-parametric model and therefore less restrictive in terms
of the assumptions the model makes about the form and distribution of the outcome: survival time.
The EC model also allows for the analysis of time-dependent variables, which are present in this
model. The two parts of the EC hazard function (see Equation 2 below) are the baseline hazard
function,
, and the exponential function which represents the independent variables in the
model.
Eqn. 2
In this equation, Xj is the time-dependent variable, denoted by the presence of (t). The
coefficients, beta and delta, are estimated using maximum likelihood techniques. The variables
included in the models for each of the three cities are outlined in the next section with an
interpretation of the results.
Survival Model Estimation Results
In this section the results of the survival analysis for each of the cities is described.
Variables of interest were explored based on evidence found in the literature. These included the
16
CIRRELT-2013-06
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
proportion of pre-1946 housing, the exposure measure, distance to the nearest park, distance to the
nearest major body of water (lake, ocean or river) distance to nearest gentrifying, or previously
gentrifying CT, and the distance to the CBD. Different interactions between variables, and between
the variables and time, were also tested. The gravity measure, Exposure, was recalibrated for each
city as the effect of transit seemed to be maximized at a different distance from the transit stations
for each urban center.
Following are the results for the three cities and a brief description of each. The number of
subjects is the number of gentrifiable census tracts, or individuals, included in the analysis, and the
number of failures is the number of census tracts that were observed to have been gentrifying at
some point during the study period.
TABLE 2: Toronto Extended-Cox Model Survival Analysis Results
Variables
Exposure
Proportion of Pre-1946 Dwellings
Proximity to Nearest Gentrifying CTs
Exposure*Time
Diagnostics
Initial Log Likelihood
Final Log Likelihood
Observations
Subjects
Failures
Coefficient
1.910
1.374
-0.119
-0.0968
Std Err
0.7893
0.4677
0.0310
0.0050
z
2.42
2.94
-3.83
-2.79
P>z
0.016
0.003
0.000
0.005
-339.987
-322.147
979
163
71
The model describing the gentrification in Toronto is summarized in Table 2. Exposure was
statistically significant with a positive coefficient meaning that as the exposure measure increases
the likelihood of a CT gentrifying also increases. It is important to understand the meaning of the
Exposure measure. In the case of Toronto, the maximum of the Exposure measure, one, is found at
a distance of 550 meters from a metro station. In order to get the odds multiplier for the variable in
this model, we raise e to the power of the coefficient for Exposure from the model. This gives us an
CIRRELT-2013-06
17
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
odds multiplier of just over 5, indicating that if a Subway station is built 550 meters away from a
census tract which previously had no access to transit, that census tract would be 5 times more
likely to gentrify as a result.
The proportion of housing built before 1946 was also statistically significant, which
reinforces findings by other analyses that consider the importance of housing stock (Walks and
Maaranen, 2008). Another variable (Proximity to Nearest Gentrifying CT) is included to answer the
question of whether the effect of gentrification is one that ‘spreads.’ This variable proved to be
statistically significant, with a negative coefficient indicating that as the distance to the nearest
already gentrifying or gentrified CT increases the likelihood of gentrification decreases.
The negative coefficient of the last variable included, an interaction of Exposure with time,
indicates that the effect of Exposure on gentrification is greatest soon after transit is implemented
and then decreases as time goes on.
TABLE 3: Montreal Extended-Cox Model Survival Analysis Results
Variables
Exposure
Exposure*Distance to CBD
Proximity to Nearest Gentrifying CT
(Exposure*Distance to CBD)*Ln of
Time
Diagnostics
Initial Log Likelihood
Final Log Likelihood
Observations
Subjects
Failures
Coefficient
1.276
-1.444
-0.133
0.335
Std Err
0.642
0.664
0.057
0.199
z
1.99
-2.18
-2.32
1.68
P>z
0.047
0.030
0.020
0.092
-439.790
-422.301
1453
220
86
The variables included in the model for Montreal were the Exposure measure, the
interaction between Exposure and the Distance to CBD, and the Proximity to Nearest Gentrifying
CT, all of which were found to be significant at or below 5% (see Table 3 above). The interacted
18
CIRRELT-2013-06
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
variable of exposure and distance to the CBD was also interacted with the natural log of time, and
was significant at 10%. The positive coefficient of Exposure tells us that as exposure increases, the
likelihood of a CT gentrifying increases. The model shows that metro stations located further from
the central business district have less of an influence on gentrification according to the interacted
variable of exposure and distance to CBD. The same may be said for the variable of Proximity to
Nearest Gentrifying CT.
Lastly the model indicates that, over time, the distance that a metro station is located away
from the CBD becomes less important, and that even metro stations located further away from the
CBD would increase the likelihood of gentrification in surrounding Census Tracts. This would
mean that as time goes on, census tracts located close to metro stations far from the CBD are
increasingly likely to undergo gentrification. This appears to be capturing the spreading of
gentrification away from the city’s center over time.
In the case of Vancouver, despite many different models being tested, only one of the
variables, distance from water, proved to be statistically significant to the onset of gentrification
with a positive coefficient indicating that as distance to water increases, the likelihood of
gentrification increases. Although this result contrasts with the hypothesis which led us to include
this variable; that the waterfront would be a type of ‘urban amenity’ with recreational facilities and
vistas, it does make sense in the context of Vancouver. First, it has been noted in the literature that
much of the gentrification in that city, especially in the West end, took place in the 1970s, such that
by the beginning of the study period much of the city was no longer considered to be gentrifiable.
Many of these gentrified areas are located along the water, in areas where boardwalks and public
parks and beaches feature prominently in the neighbourhood. Of the ‘gentrifiable’ census tracts
considered for analysis in this paper, almost the entire shoreline is used for industrial uses,
removing the ‘amenity’ that the waterfront may otherwise offer to nearby residents. Nevertheless,
CIRRELT-2013-06
19
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
this model should not be interpreted as a stand-alone explanation of gentrification in Vancouver, but
indicates that in the case of Vancouver, more investigation is needed.
The difference in Vancouver in the relationship between transit and gentrification, or the
lack thereof, is consistent with recent findings on the city of Vancouver, which demonstrate that
poverty is actually spreading along the Skytrain lines, rather than gentrification as seems to be the
case in other cities, such as Toronto (Ley and Lynch, 2012). Due to all of these factors, it is
understandable that the results should differ significantly from those of the other cities.
Despite the attention paid to the importance of urban amenity in the literature, and the
attempts of the authors to describe this relationship with the statistical model, using urban parks and
proximity to water, ultimately these indicators were no more helpful in describing the process of
gentrification in Canadian urban centers.
DISCUSSION AND CONCLUSION
The difference in results from the three cities is not all together unexpected considering
their varying contexts.
Vancouver, the city which diverges most from any noticeable trend is the one with the
youngest transit system, therefore the shortest study period. Despite the results of the model in
Vancouver, there is still an identifiable pattern in the analysis of Toronto and Montreal, as well as a
statistically significant relationship between transit and the onset of gentrification. The importance
of the addition of the variable Proximity to Gentrifying CTs must not be underestimated. With the
addition of this variable the model shows that while proximity to gentrifying CTs is important, the
presence of rapid rail transit remains statistically significant.
Though this paper provides an important foundation for future studies, there are still many
questions about the relationship between transit and gentrification and what other urban elements
20
CIRRELT-2013-06
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
may be contributing to this process. It will be interesting to watch, as new censuses become
available, how the trend evolves over time in different urban centers as well as expanding the
studies to encompass smaller, regionally important urban centers and other forms of transit such as
Bus Rapid Transit.
Finally, being able to firmly conclude that rail transit has a significant impact on
gentrification in two of the cities, the study should be used to inform planners and researchers about
the many effects of the implementation of transit, and how best to mitigate the negative effects of
gentrification and displacement, which may occur as a result of increased accessibility to transit.
This study should be seen as an innovative and applicable starting point for research looking into
transformations experienced in part or in whole as a result of urban transit coming into operation.
ACKNOWLEDGEMENTS
This research has been funded through the Canadian Social Sciences and Humanities Research
Council (SSHRC) Canada Research Chair’s program as well as the Joseph-Armand Bombardier
Canada Graduate Scholarships Program.
In addition, the authors would like to extend a special thanks to: Craig Townsend, Alex Guindon
and Robert Tomaszewski at Concordia University, Nathalie Vachon and Dany Fougères at the
Institut National de Recherche Scientifique (INRS), and Marcel Fortin at the University of
Toronto’s Computing in the Humanities and Social Sciences (CHASS) as well as Richard
Maaranen and the Centre for Urban and Community Studies for providing access to digitized
census data and digital boundary files. Their help was essential to the progress of this research.
Special thanks to Chris Harding and Arash Mostofi for their support.
CIRRELT-2013-06
21
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
REFERENCES
Atkinson, R., and G. Bridge. (2005) Introduction: in R. Atkinson and G. Bridge (Eds.)
Gentrification in a Global Context; the new urban colonialism. New York: Routledge, 2005.
Atkinson-Palombo, C. (2010) Comparing the Capitalisation Benefits of Light-ril Transit and
Overlay Zoning for Single-family Houses and Condos by Neighbourhood Type in Metropolitan
Pheonix, Arizona, Urban Studies, 47(11):pp. 2409-2426.
Patterson, Z., and Annelise Grube-Cavers, (2011) Metros and Gentrification in Montreal - A
Survival Analysis Approach. Washington, D.C.: Presented at Transportation Research Board 91st
Annual Meeting.
Boyle, M. (1995) Still top of our agenda?: Neil Smith and the reconciliation of capital and
consumer approaches to gentrification. Scottish Geographical Magazine 111: pp. 119-123.
Cervero, R., M. Bernick, and J. Gilbert. (1994) Market Opportunities and Barriers to Transit-Based
Development in California, University of California Transportation Center.
Clairoux, B. (2001) Le Métro de Montréal; 35 ans déjà. Montréal: Hurtubise HMH Ltée.
de Dios Ortúzar, J., and L.G. Willumsen. (2001) Modelling Transport; third edition. Chichester:
John Wiley and Sons, Ltd.
Filion, P. (1991) The Gentrification - social structure dialectic: A Toronto case study. International
Journal of Urban and Regional Research 15(4): pp. 552-574.
Freeman, L. (2005) Displacement or Succession? Residential Mobility in Gentrifying
Neighborhoods, Urban Affairs Review, 40(4):pp. 463-491.
22
CIRRELT-2013-06
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
Glass, R. (1964) Aspects of Change, in: R. Glass (Ed.) London; Aspects of Change, pp. xiv-xxxvi.
London: MacGibbon & Kee Ltd.
Hackworth, J., and J. Rekers. (2005) Ethnic Identity, Place Marketing, and Gentrification in
Toronto, Research Paper 203. Centre for Urban and Community Studies: pp. 1-22.
Helms, A.C. (2003) Understanding gentrification: an empirical analysis of the determinants of
urban housing renovation, Journal of Urban Economics, 54(3):pp. 474-498.
Hess, D.B., and T.M. Almeida. (2007) Impact of Proximityy to Light Rail Rapid Transit on Stationarea Property Values in Buffalo, Urban Studies, 44(5-6): pp.1041-1068.
Kahn, M. (2007) Gentrification Trends in New Transit-Oriented Communities: Evidence from 14
Cities that Expanded and Built Rail Transit Systems, Real Estate Economics, 35(2):pp. 155-182
Kennedy, P. (2005) A Guide to Econometrics; fifth edition. Bodmin: The MIT Press.
Kleinbaum, D.G., and M. Klein. (2005) A Self-Learning Text; Second edition. New York: Springer.
Ley, D. (1986) Alternative Explanations for Inner-City Gentrfication: A Canadian Assessment.
Annals of the Association of American Geographers, 76 (4) pp: 521-535.
Ley, D., and N.A. Lynch. (2012) Divisions and Disparities: Socio-Spatial Income Polarization in
Greater Vancouver, 1970-2005. Neighbourhood Change and Building Inclusive Communities from
Within.
http://neighbourhoodchange.ca/documents/2012/08/summary-version-divisions-and-disparitiessocio-spatial-income-polarization-in-greater-vancouver-1970-2005-by-david-ley-nicholas-lynch.pdf
(accessed November 23, 2012).
Lin, J. (2002 ) Gentrification and Transit in Northwest Chicago, Transportation Quarterly, 56(4):pp.
235-263.
CIRRELT-2013-06
23
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
Newman, K., and E,K. Wyly. (2006) The Right to Stay Put, Revisited: Gentrification and
Resistance to Displacement in New York City. Urban Studies. 43(23): pp. 23-57.
Maddala, G. S. (1983) Limited-Dependent and Qualitative Variables in Econometrics. Cambridge:
Cambridge University Press.
Metro Vancouver. (n.d) Local Governments of Metro Vancouver. Metro Vancouver.
http://www.metrovancouver.org/about/maps/Maps/Municipalities.pdf (accessed July 25th, 2012).
Murdie, R., and C. Texeira. (2011) The Impact of Gentrification on Ethnic Neighbourhoods in
Toronto: A Case Study of Little Portugal, Urban Studies, 48(1): pp. 61-83.
Pollack, S.B., B. Bluestone, and C. Billingham. (2010) Demographic Change, Diversity and
Displacement in Newly Transit-Rich Neighborhoods. Transportation Research Board 2011 Annual
Meeting. Washington, DC.
Rose, D. (2004) Discourses and experiences of social mix in gentrifying neighbourhoods: a
Montreal case study. Canadian Journal of Urban Research. 10(3): pp. 229-245.
Skaburskis, A, and D. Mok. (2006) Chapter 5: Cities as Land Markets: In T. Bunting and P. Filion
(Eds.) Canadian Cities in Transition; Local Through Global Perspectives, pp. 86-101. Toronto:
Oxford University Press, 2006.
Slater, T. (2004) Municipally managed gentrification in South Parkdale, Toronto, The Canadian
Geographer, 48(3):pp. 303-325.
Slater, T. (2006) The Eviction of Critical Perspectives from Gentrification Research, International
Journal of Urban and Regional Research, 30(4): pp. 737-757.
Smith, N., and P. Williams (1986) Alternatives to orthodoxy: invitation to a debate: in N. Smith and
P. Williams (Eds.) Gentrification of the City: pp.1-10. London: Allen & Unwin.
24
CIRRELT-2013-06
Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres - A Survival Analysis Approach
Statistics Canada. (2006) Census Statistics. Ottawa: Statistics Canada, 2006. —. Census Tract:
detailed definition. 23rd April 2007.
http://geodepot.statcan.ca/2006/Reference/COGG/LongDescription_e.jsp?GEO_LEVEL=12&REF
CODE=10&TYPE=L (accessed July 25th, 2012).
Statistics Canada. (2007) Statistics Canada Census Statistics; 2006. Ottawa: Statistics Canada
Toronto Transit Commision. TTC Milestones. 2012.
http://www.ttc.ca/About_the_TTC/History/Milestones.jsp (accessed July 25th, 2012).
Walks, A., and R. Maaranen. (2008) The Timing, Patterning and Forms of Community
Gentrification and Neighbourhood Change in Montreal, Toronto and Vancouver, 1961-2001.
Toronto: Center for Urban and Community Studies.
Washington, S. P., M. G. Karlaftis, and F. L. Mannering. (2003) Statistical and Econometric
Methods for Transportation Data Analysis. Florida: CRC Press.
CIRRELT-2013-06
25